Erider torque voltage output consistency issue

Philtek

1 mW
Joined
Mar 22, 2015
Messages
17
Hello ES!

I have an Erider bottom bracket that is outputting inconsistent torque voltages. Perhaps one of you have seen similar issues and can comment.

Issue: Voltage output from torque sensor stops modulating during riding. This usually occurs within a 5-10 minutes of riding. Power cycling the system resets the sensor and it works normally again.

Observations:

36 pole Erider, 48 volt system, CA V3(tried a few FW same result), Phaserunner, All Axle STD wind (same condition observed with GMAC).

Confirmed 11.56v at Erider.

Confirmed the CA V3 is reading the voltage correctly: I thought the CA firmware might have a bug but I have confirmed the torque voltage output from the Erider at some point stops modulating and sits at 1.56 volts no matter how hard I crank the pedals. I used a multimeter on the handlebars while riding and confirmed the flatlining of output voltage from the Erider at 1.56 volts when the issue appears. Only way to resolve is to power cycle the system which immediately resolves the issue.

Tried swapping out the Erider sensor shell with a second unit: I am still observing the same behavior even with a second Erider sensor shell in the same installation. I only swapped the sensor shell (left side sensor cup) , not the entire shaft or right side cup. I’ve swapped the sensor wiring as well as I originally suspects a simple wiring issue.

Seems to occur most often after heavy pedaling, or shifting under load: When the issue appears, I stop and observe the torque voltage the CA is reading in the setup menu. Most of the time when the issue crops up the sensor will sit at 1.56 volts no matter how hard I crank. Sometimes it modulates only slightly while really cranking on the pedals and other times not at all. A power cycle resets it completely and normal operation returns every time so I suspect an issue with the IC in the sensor itself. Maybe an out of range reading on the sensor is causing the sensor to crash or recalibrate to a much higher load threshold.

I have read that the Erider can be susceptible to EMF noise. Has anyone come across of a condition as I’m describing being caused by EMF noise? Is there a common installation fault (too tight cups or lock ring) that could has caused issues with the sensor? I have tried moving the sensor wiring away from any other cables or the frame but the issue has persisted. My next move will be installing a small cap on the 12v-gnd at the pas JST and seeing if that helps.

Thanks in advance.
 
Update: Tightened the cups and noticed some improvement

I noticed the dive side cup came off easily so I suspect during installation I must have forgotten to tighten down properly (20-30Nm). I guess the spindle must have been slightly out of alignment relative to the sensor cup on the non drive side. A test ride showed relatively stable performance however the problem has returned in subsequent rides.

Next step would be scoping out the sensor while riding and seeing what I find. I would think I could observe some noise but I'm not sure to be honest.
 
I took some scope captures today while poking around and found the 12 volt to ground supply to the Erider was stable, but the torque voltage to ground had some noise present. I took readings at the sensor side JST connector between the juliet connector and the CA. These captures were made at idle without the motor running. Running and loading the motor didn't seen to affect the noise observed.

DS1Z_QuickPrint4.png
The captures above and below show a short double pulse about 8.8 us apart or 115~ kHz. The transient spike peaks around 4 volts while the sensor is outputting its torque voltage (1.6 volts or so in this capture)
DS1Z_QuickPrint5.png
The photo below shows a reading from the same point after .01 uF capacitor added between ground and torque output
DS1Z_QuickPrint6.png

A test ride after crimping the capacitor in place with the torque signal and ground pins at the JST yielded promising results with consistent torque readings but I'll have to test ride further to see if the issue is fully resolved as it tends to creep in 10-15+ minutes after boot up. I also loosed the drive side cup a touch as I reckon the cups are tightening against each other inside the bottom bracket, and I thought perhaps if they are over tight this could cause a strain issue but I 'm really not sure how these sensors do what they do exactly.

I'll report back after a few days.
 
Philtek said:
Hello ES!

I have an Erider bottom bracket that is outputting inconsistent torque voltages. Perhaps one of you have seen similar issues and can comment.

Issue: Voltage output from torque sensor stops modulating during riding. This usually occurs within a 5-10 minutes of riding. Power cycling the system resets the sensor and it works normally again.

I've noticed exactly the same problem! Didn't happen at the beginning, but for two days now I constantly have to restart the system to rid the issue, just like you describe! One thing I noticed is that sometimes the problem can be trigger when hitting a pothole hard. My first thought was that the TQ sensor outputs a voltage peak past the "fault level" set in the CA, triggering the CA to shut it off. But the noise issues you describe seems more like a systemic internal design error of the sensor.

Very interested into finding a solution. Pictures of your setup would be welcome!

Also, you might want to send your finding to info@ebikes.ca. I'm pretty sure this problem is not isolated, and they should be interested into a solution as well, and have access to more user feedback.
 
Hi Philtek & qwerkus,

I've got an Erider T9, CA3 & Baserunner and I also experience occasional need to reboot the whole system after hitting large bumps, or potholes. This happens only occasionally for me because I usually try my utmost to avoid all potholes. I originally thought this was the Baserunner throwing an 'instantaneous phase overcurrent' fault and needing to be rebooted to clear the fault.

Next time this happens, I will see if I can pull over and have a look at the TQ voltage in the CA3 setup screen and see if the T9 is still modulating the TQ voltage before I reset it.

I actually have both the older 18 pole and the newer 36 pole versions of the T9 so I can throw them both on the scope for you.

Do you see glitches in the TQ line immediately after power up?
Or does it only start to gltch 10-15mins after power up?

thanks,
Oli.
 
Hi,

I scoped both my 36 pole and 18 pole Erider T9s. The measurements were taken with a 10x passive probe between the TQ line and the GND of the JST connector. The 36 pole was tested on the bench, and the 18 pole was tested while fitted to the bike. Each was powered by the CA3 on the bike. I saw no significant difference between the traces of the two versions at these particular zoom levels and I couldn't see any noise of the magnitude that you are seeing.

The traces below are from the 36 pole. They are zoomed to the same time base and voltage as you used for comparison.

#1 Same as your first image H: 2us V: 2.00V
36 pole 2us 2v.png

#2 Same as your second image H: 50ns V: 2.00V
36 pole 50ns 2v.png

I did search for your noise using various different triggers and I couldn't find anything like what you are seeing, but when I zoomed right in to 10x where you were, at 200mv/div I started to see some noise but this was 10x smaller than what you are seeing at about 300mV P-P vs. your 4V P-P.

#3 Zoomed in 10x to H: 2us, V: 200mV
36 pole 2us 200mv.png

Thanks,
Oli.
 
My probes were on 10x as well and this makes sense as I was reading 1.6 ish volts from the sensor according to the scope. I rode around town today and although improved, I did still have to reset the system several times to get normal human power readings back on the CA, and therefore normal torque assist power output.

I seemed to notice sudden hard braking/regen caused the sensor to suddenly drop its torque voltage output. Of course when I try to replicate it, it seems to behave well. It is a perplexing issue for sure and likely due to emf noise or a transient spike somewhere, I'm just not sure where.

Do you see glitches in the TQ line immediately after power up?
The noise pictured in my scope captures was immediate, with the bike on but sitting on my work stand. Powering the motor on or off did not change this observed noise.
Or does it only start to gltch 10-15mins after power up?
It seems to start soon as the system is power on. My guess is its noise from the Phaserunner but it's just a theory.

Below are photos of the install as it is now. I have an PAS extension cable in place, however I have tried removing the extension, moving the torque sensor cabling completely away from the frame, away from the battery, away from the CA bundle etc without a meaningful improvement of the condition. I did once observe the human watts jumping around all over the place, like -500 200,-800, etc but when I got home and got it back on the stand this particular fault condition seemed to have disappeared on it's own.

IMG_20210702_160052612.jpgIMG_20210702_160058394.jpgIMG_20210702_160104199.jpgIMG_20210702_160123991.jpgIMG_20210702_160138483.jpgIMG_20210702_160145685.jpgIMG_20210702_160148373.jpg
 
Philtek said:

Lol, I have the exact same bike (rebranded bbf) with the same issue. What motor do you use ? I have a front FH212 DD hub. If you route the tq cable straight down the steering tube and than directly to the bb, you gain around 50cm cable length.
I replaced the front fork with a suspended one to improve ride quality though I does not help with the tq sensor issue. I can confirm that the problem shows up after regen braking and hitting a pothole. Sometimes hard pedaling seems to be enough though, which really sucks when you have to restart the system in the mids a the climb for an instance!
 
Qwerkus: I did originally have the cable routed the way you described. I have tried routing the cable differently in an effort to address the issue but so far, nothing has worked.
 
Here the helpful reply from ebikes.ca (Peter F.)

There are two main failure modes I've been seeing in Erider torque sensors. I'm just coming to understand these issues, and this is by no means definitive.

One, where the torque signal voltage freezes at the neutral offset, and is cured by power cycling, seems to be the result of the MC in the spindle crashing, likely caused by noise. In these cases the Erider is generally fine, and replacing it does not resolve the issue, being the result of EMI on the frame. My suggestion in these cases is to return the Erider, and fall back on a simpler PAS.

Two, where the sensor offset drifts and loses sensitivity. Many users who report this also report having to use force to install the sensor. The cause is not well understood, but regardless, the sensor can generally still be made to function. Requiring the neutral offset to be rezeroed. And the scale factor to be increased to account for the decrease in sensitivity. If accuracy is desired, one can use a torque wrench on the spindle to apply a known torque. Scale factor (Nm/V) equals the Nm applied, divided by the change in signal voltage. I have simply tweaked the scale factor to bring hW into a sensible range The drift seems to be a single discrete change, rather than continuous drift, making this recalibration a useful solution in many cases.

My first t9 sensor definitely died. Only weird tq output signal readings (from negative to +10V) so I'm not sure it's a bug - rather a defective unit.
 
Wow, so happy to find others are having the same issue I am. It would be amazing if we could figure this out, as my bike is otherwise perfect and I love riding it.

I'm on my third ERider TS BB. The first two were the previous-gen 18-pole versions, the current one is the 36-pole version. I've had issues with all of them. The first one appeared to be broken out of the box, with a no-torque voltage of around 2.5 V and with a very limited torque range (it would basically report an additional 10-20 Nm and then max out). The second one worked for around 500 mi and generally worked well, except for two failure modes: If I stopped and kept my foot on the pedal for more than a few minutes, the TS's internal zero-point calibration would get messed up and it would effectively stop reporting any torque. The only solution to this was to wait. Now and then, I seem to have also run into the problem of hitting a big bump and the TS flatlining, but I did not have any issues with the e-brake (modulated using the throttle) causing the TS to stop responding. After around 500 mi, this TS suddenly started reporting around 2.5 V even with no torque on it, so I had to return it to Grin for an exchange. It happened overnight, from one day to the next. I have no idea what caused it to happen. I remember, though, that this BB had a pretty substantial dab of epoxy on the spindle, so maybe that was actually rubbing against the interior of the shell?

With my third TS, I had to upgrade the CA firmware to 3.15b3 in order to get 36-pole averaging. Now, more frequently than before, the ERider will flatline and the CA will report no pedal torque. However, I'm not convinced this is a problem with the ERider, because if I go into the CA setup mode without power-cycling it, the CA reports both torque and pedal motion. If I then exit the setup mode, the CA again sees human input and PAS works as before. It feels like this happens more often if I: 1. Pedal hard, or 2. Use the throttle to modulate the e-brake. It almost feels like there is some kind of crosstalk between the throttle input and the PAS input into the CA that ends up corrupting the CA's state. Or, maybe it's electrical; I don't know. One weird phenomenon I've seen in this state is that if I pedal slowly, the CA will actually report negative human watts (HW).

I wish I could take a look at the CA source code to see what the problem could be.
 
I suspect you're right - I'm having problems distinct to the 3.15b3 beta firmware that don't happen when I've tried flashing the CA to a non-beta firmware. Throttle modulated regen sometimes kills motor outputs - all of the inputs (rpm, speed, etc.) register fine. Sometimes torque sensor driven output doesn't work, sometimes throttle output doesn't work, sometimes brake cutoff doesn't work, and usually regen is gone. Twice during my last ride it crashed to a state where throttle input works and brake cutoff doesn't. That's a safety problem, so I'm reverting to non-beta firmware until Grin gets the CA figured out. I'd also be exchanging this bottom bracket as well, but it's the only standard size option they offer, so that's not an option.

The non-beta CA firmware still faceplants sometimes when there's a lot going on. 36 poles + very high pedaling cadence will also cause problems, for example. There have been comments in the past that the pole count limit was set at 32 due to microprocessor limitations. The failure modes I've encountered and heard of are consistent with processor constraint collisions.
 
Well you can always halve the pas signals to 18. Still plenty for a near-immediate startup, and never causes problems. I don't think those problem reports are due to the ca limitations; e rider has serious quality issues. A quality tdcm unit costs like twice as much...
 
I am starting to look for a kit for my GF's bike and it seemed like these rear geared hub setups with torque BB would be perfect. My bikes are mid drive and while they work fine I can see where issues could arise. I'd like her bike to be bullet proof. If the E-rider torque BBs aren't good can you recommend others that are better? I see that the Grin guys sell Sempu for about the same price. Are they better and not subject to the drops outs and such?

qwerkus said:
Well you can always halve the pas signals to 18. Still plenty for a near-immediate startup, and never causes problems. I don't think those problem reports are due to the ca limitations; e rider has serious quality issues. A quality tdcm unit costs like twice as much...
 
raylo32 said:
I am starting to look for a kit for my GF's bike and it seemed like these rear geared hub setups with torque BB would be perfect. My bikes are mid drive and while they work fine I can see where issues could arise. I'd like her bike to be bullet proof. If the E-rider torque BBs aren't good can you recommend others that are better? I see that the Grin guys sell Sempu for about the same price. Are they better and not subject to the drops outs and such?

Can't tell from my own experience, but from what I've seen, the tdcm has a much better build quality: https://www.heinzmann-electric-motors.com/downloads/directpower-system/send/22-directpower/107-btts-sensor-bedienungsanleitung

I'd consider things twice though: do you really need a torque sensor ? This thing works like an amazing way to controller speed / power on flat lands where you'd pedal anyway. Here in the swiss mountains, it really sucks: you have to pedal quite hard all the time to get appropriate assistance when going uphill. I found out that a PAS with a continuous assistance setup (potentiometer) works MUCH better.
 
Both my mid drive e-bikes have torque sensors that work perfectly. Climbing 10-15% grades can be a real grind or easy peasy depending on assist level. I can't even imagine riding a PAS bike... unless you just don't like the sensation of pedaling and using that to control your speed. Sorry to hear these don't work better.

qwerkus said:
Can't tell from my own experience, but from what I've seen, the tdcm has a much better build quality: https://www.heinzmann-electric-motors.com/downloads/directpower-system/send/22-directpower/107-btts-sensor-bedienungsanleitung

I'd consider things twice though: do you really need a torque sensor ? This thing works like an amazing way to controller speed / power on flat lands where you'd pedal anyway. Here in the swiss mountains, it really sucks: you have to pedal quite hard all the time to get appropriate assistance when going uphill. I found out that a PAS with a continuous assistance setup (potentiometer) works MUCH better.
 
Update: I think I can reproduce the bug when I e-brake and pedal at the same time. Otherwise, after updating the CA firmware again and wiping the calibration, I seem to be able to use the e-brake even with throttle modulation without triggering the bug.

I first tried to downgrade the CA's firmware from 3.15b3 to 3.14 to see if that would fix things. I input my settings as they were before, except, I programmed the CA to expect an 18-pole (not 36-pole) TS with a scale factor of 35 Nm/V (instead of 70 Nm/V). For whatever reason, this appears to have caused PAS to totally stop working. I made one loop around the block and the PAS never kicked in. It's possible that after reprogramming the CA, the HW multiplier was just set at zero, though, I forget if I checked for this. Also, I did not check to see if it registered anything from the torque sensor at all by going into the setup menu.

In any case, I then reflashed the CA to 3.15b3 and I had the CA programming software wipe the calibration settings. After re-entering my settings into the CA, including setting a 36-pole TS with the correct 70 Nm/V scale factor, when I powered up the CA on the bike, it showed -3 W usage at standstill. Well, that's obviously not correct, so, I went into the CA setup menus and zeroed-out the power measurement at idle. I suspect this last step might've been the key, because I can now use the PAS pretty well and the bug where PAS disappears doesn't seem to get triggered unless I'm also pedaling while the e-brake is engaged. If I tap the brakes while riding without pedaling, I can sometimes get this bug to go away.

So, not a total fix, but maybe one step closer to figuring out what's happening. It still looks like there's an interaction between the PAS input and the e-brake that triggers a bug in the CA that causes it to register PAS input as contributing negative human watts even after the e-brake is released. Of course, PAS input should never be treated as negative.
 
matmat are you seeing large variations in human watts? I observed this only one time and didnt really do anything other than moving wires around and the issue has not returned.

My ongoing issue is loss of torque output voltage from the torque sensor itself. I have observed the loss of signal output as I ride with a multi meter on the handlebars. I suspect the small PCB on the spindle is resetting but I don't know why. A power supply issue through the transfer coils between cup and spindle? Maybe noise interfering with actual load amp circuit? Hard to say as I cant scope the spindle in action for obvious reasons...

I'm hoping to find a solution because I really want to use torque sensors that are easy to install for my upcoming builds but there is some issue here. I'm trying to source some torque sensor compatible controllers from overseas to test with perhaps I will have to go that route.
 
The spindle is pictured below. My guess is the coil is used for powering the IC and transmitting data.

IMG_20210728_215213535.jpgIMG_20210728_215235829.jpg
 
Philtek said:
My guess is the coil is used for powering the IC and transmitting data.
Thank you for the photos! So there is only one coil? The old sempus had two coil pairs, one for power transmission and one for data transmission:

index.php


Some years ago, I built a sensor prototype for measuring the torque from the crank or from the chainwheel. It used only one coil pair for power and data also, the torque information was transmitted as a pulse frequency modulated on the supply current.

index.php


regards
stancecoke

p.s.
there is a photo of the transmitter side in the german forum also:
index.php
 
Wow stancecoke thanks for sharing that info. It does indeed appear to be a single coil on the Erider. I tried swapping the phaserunner for a grinfineon but same behavior so far... not a surprise but going through swapping components in case I observe a change.

I wonder if the issue is more pronounced on an aluminum frame? I'm sure we've all seen the magnet in aluminum pipe trick- how the magnet slowly falls down the pipe. Maybe the aluminum bottom bracket is absorbing some of the magnetic field that the cups is producing to communicate with the spindle? No idea just a thought I had.

Is anyone with a steel bike experiencing the same issue with their Erider?
 
Stancecoke, thanks for the great pictures!

Philtek, I have not seen a large variation in human watts. The typical problem I see with the CA 3.15b3 software is that if I use the pedals while e-braking or while using the throttle (new observation!), the CA stops registering any contribution from the pedals (human watts drops to zero) until I either power cycle it or go in/out of the setup menu.

I'm interested in your observation that you can measure the E-Rider's output using a voltmeter and see it flatline under some circumstances. I'll try that out, too, just to verify that it really is a problem with the CA that I'm seeing.

The idea of EMI from the frame is interesting. IIRC, the example of the magnet falling slowly through a tube is a demonstration of Lenz' law, where the falling magnet causes a varying magnetic field, which induces an EMF in the tube, which induces a current that's a function of the conductivity of the tube, which then creates a magnetic field in the opposite direction of the original field. So, among other things, the strength of the induced magnetic field is a function of the conductivity of the material:

https://www.tibtech.com/conductivite.php

According to the above, the conductivity of copper is about 1.6x of aluminum, while the conductivity of steel is about 25% of aluminum. All that to say, I could see it that the aluminum BB shells are actually creating back EMF that might interfere with their operation, especially if it's at a resonant frequency, while the steel BB shells might not.

I'm trying to think of some way that you might be able to block the back EMF from the BB shell... and I can't. I think the torque sensor's shell should be doing that, but if it's steel, then, it's not going to do as good a job as aluminum would. Maybe try wrapping the E-Rider in several layers of aluminum foil taped tight?
 
matmat said:
Maybe try wrapping the E-Rider in several layers of aluminum foil taped tight?
Ferrite sheets are very effective for electromagnetic shielding and field concentration. Sadly the flexible ones are quite expensive. :shock: The solid ones are much cheaper.
They are used in wireless charging modules, the power transmission from the housing to the shaft of the torquesensor is nothing else than wireless charging...

o14789v209%20Family_WE-FSFS.jpg

He6a27dfd93bf4743bfc340807a03c42f1.jpg


regards
stancecoke
 
Some updates.

First, stancecoke, do you see this kind of problem with the open source KT controllers? I mean, if you hook up an E-Rider to one of the KT controllers, will it randomly stop working?

I exchanged some emails with Peter F at Grin about the situation, but he was not very optimistic. My understanding is that E-Rider + GMAC + CA malfunctioning due to EMI is one of the most common kinds of failure. Peter suggested that shielding, RF chokes, and cable rerouting may help, but was unsure of the root cause of the problem (could be EMI in the wires, radiated EMI, or both) and kept coming back to letting me know that the ultimate solution was probably going to be to drop down to a less sophisticated torque sensing BB that did not use wireless power and/or data transfer and did not have these issues (the NTCE sensor).

Meanwhile, I stumbled on this post: https://endless-sphere.com/forums/viewtopic.php?f=2&t=107750 Same problem as we're discussing in this thread, apparently more-or-less fixed using some electronics and electrical components, but no shielding. Interesting.

In trying to figure this out, I was browsing through the unofficial CA guide, and in it, there is a table of how much current accessories can draw at various battery voltages. Long story short, at ~52 V, it's only about 40 mA, but the E-Rider can draw as much as 400 mA according to its documentation! That sounded like the E-Rider could be causing a voltage drop in the CA and would explain why we see the problem almost exclusively when the e-brake or throttle (that is, other accessories) are being used. So, I hacked together a 12 V supply to power the E-Rider, while also connecting its ground to the CA and E-Rider grounds. The CA was still powered by the battery. This immediately and reliably fixed the problem of the PAS malfunctioning when using the e-brake, but not when using the throttle. Using the throttle would still cause the PAS to malfunction.

I next installed some simple RF chokes: Between the GMAC and the Phaserunner, between the Phaserunner and the CA, between the E-Rider and the CA, and between the CA and the throttle. These helped the problem somewhat, but have not been a sure-fire fix.

Where to go from here? Peter F tells me that the Sempu BB's are just as prone to this behavior and they require a hole in the BB shell, so, that's probably out.

I'm still not convinced that the problem is with the E-Rider and not with the CA. Now that the E-Rider is powered independently, it's clear that it doesn't get reset or power cycled when I turn the CA on/off, and yet, turning the CA on/off will still fix the PAS malfunctioning. It stands to reason that if the problem were with the E-Rider, then power cycling the CA would not fix the issue, and yet, it does. My suspicion is that the problem is fundamentally with the CA.

Next, I will probably try to power the CA, the throttle, and the E-Rider all from external DCDC converters (12 V for the E-Rider and the CA, 5 V for the throttle). I may also try to put some capacitors on the E-Rider power supply line (as suggested in the liked ES post). I'm not keen on stepping down to an NTCE BB, since they only register torque on the left pedal and also have many fewer poles.

If the problem were truly with the Phaserunner and/or GMAC (as suggested by Peter), I would consider swapping out the GMAC for the RH212, but that's a lot of time and money for something that may/not work. Another possibility is switching over to a KT controller with the open source firmware, but I could just be trading one set of bugs for another. :lol:
 
Back
Top